Non-stationary link inference and localization in communication networks

Existing network link estimation methods generally assume that the network link status in the measurement is stationary, but this assumption is not always true in the real network. Thus they cannot provide desired estimation accuracies. To address the problem, in this paper, we propose a new methodology, which can accurately infer the packet loss rates of all links in the network and locate the non-stationary links. Through software simulation, we compare our method with a former inference algorithm (LIA). Experimental results show that the new algorithm can provide higher inference accuracy within the same computing time.